[m-dev.] ODBC interface

[Simon Taylor]

 > 
> > This can be left up to the caller as well. It isn't a problem
> > for the ODBC interface since the threaded odbc__states ensure
> > that an exception can't happen within builtin_aggregate.
> 
> I don't know anything about the ODBC interface or exceptions, so I may be
> way off on this, but I don't see how threading odbc_states ensures that an
> exception can't happen within builtin_aggregate.  You could pass an
> odbc_state as (part of) the accumulator.  This might be useful to fill a
> database with the solutions to some nondet goal.

No, I was way off, I was thinking of solutions, not aggregatge. It still 
isn't a problem for the ODBC interface since that doesn't work without 
conservative GC anyway. 

Simon.